Blind deconvolution of turbulence-degraded images using natural PSF priors

Deconvolution of images taken through atmospheric turbulence often requires regularization in order to prevent the restoration algorithm over-fitting the noisy observation. For this purpose many object priors have been proposed but their utility might be limited to one class of real objects. Optical effects of atmospheric turbulence are well understood and therefore priors on the point-spread function form a viable alternative. We show their usefulness.